[ https://issues.apache.org/jira/browse/HBASE-23770?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Michael Stack updated HBASE-23770: ---------------------------------- Attachment: Screen Shot 2020-01-29 at 9.45.58 PM.png > [Flakey Tests] TestRegionReplicasWithRestartScenarios#testWhenRestart > --------------------------------------------------------------------- > > Key: HBASE-23770 > URL: https://issues.apache.org/jira/browse/HBASE-23770 > Project: HBase > Issue Type: Bug > Components: flakies > Reporter: Michael Stack > Priority: Major > Attachments: > 0001-HBASE-23770-Flakey-Tests-TestRegionReplicasWithResta.patch, Screen Shot > 2020-01-29 at 9.45.58 PM.png > > > Fails about 35% of the time in the GCE build. Let me attach a picture from > current flakies dashboard for branch-2. > The test starts a cluster of three RS w/ 3 region replicas. It then stops a > server, starts a new one, and then expects that the remaining three nodes do > not have instances where two region replicas have landed on a single server. > It fails sporadically (reproducible locally) because when the SCP runs its > assign, sometimes timing has it so Master knows of two servers only. Making > the new start before the old one is stopped (instead of other way around) > seems to fix the test -- there'll be three servers up when SCP runs. -- This message was sent by Atlassian Jira (v8.3.4#803005)